Hi all,
rather than "interpolation" it is more correct "filtering";
as with analog signals, to reduce 88 to 44 it is required to perform
filtering, that is removing all frequency component higher than half
the final sample rate. "Throwing" a sample every two is correct, but
AFTER the filtering. Dithering could be an option to redistribute in
the frequency spectrum the digitization noise.
The filtering of either 88 or 96 is almost equivalent in terms of
computations, but reducing from 88 to 44 is much simpler and
straightforward than reducing 96 to 44. "Probably" the process is
also cleaner, with less round-off errors.
Of course each software uses its own resampling algorithms, each with
its own features.
Gianni
At 21.41 14/02/2009, you wrote:
>Thanks, Lou, for pointing out my mistake. I wrote
>
>"Making the original recording with 88.2 kHz means the
>reduction to 44.1 is simpler (just throw out every other sample) than,
>say, the conversion from 96 kHz to 44.1 kHz."
>
>This is incorrect. When doing sample rate conversion from 88.2 to
>44.1, the process that sounds best is not a simple subtraction of half
>of the data. The conversion includes dithering and interpolation,
>too, which makes the process very similar to that used to convert from
>96K samples to 44.1K samples.
>
>I also wrote, "Some people think this [recording at 88.2 kHz]
>helps preserve the sound of the original recording." This part
>is true, some people do prefer 88.2, which is why Olympus added
>this sample rate to the capabilities of the LS-10.
>
>--oryoki
>
>
>
>Nessun virus nel messaggio in arrivo.
>Controllato da AVG - www.avg.com
>Versione: 8.0.237 / Database dei virus: 270.10.23/1952 - Data di
>rilascio: 02/13/09 18:29:00
---------------------------------------------------------------------------=
----------------------------
Gianni Pavan
Centro Interdisciplinare di Bioacustica e Ricerche Ambientali
Universita' degli Studi di Pavia, Via Taramelli 24, 27100 PAVIA, ITALIA
Phone +39-0382-987874 Fax +39-02-700-32921
Email
Web http://www.unipv.it/cibra
---------------------------------------------------------------------------=
-----------------------------
|